Pre-service technology teachers demonstrate modest overall advantages in AI literacy compared to secondary technical students, with competency-level gains concentrated in understanding intelligence and programmability, yet no between-group differences emerge in critical AI literacy, revealing that exposure alone is insufficient without structured, performance-based curricula embedding data practices, ethics/governance, and human–AI design across both educational sectors alongside targeted interventions to close persistent gender gaps.
Objective: This two-phase study aimed to (1) map dominant themes and intellectual bases of AI literacy in technology/engineering education through bibliometric analysis and (2) empirically compare AI literacy between pre-service technology teachers (n=68) and secondary technical students (n=77) to inform curriculum design. Four research questions guided the work: RQ1 examined thematic evolution; RQ2 assessed overall literacy differences; RQ3 identified competency-specific gaps; and RQ4 compared critical AI literacy between cohorts.
Methods: Phase 1 queried Web of Science (2015–2025) for AI/data/critical literacy in educational contexts, yielding 1,259 documents across 587 sources. R/bibliometrix generated thematic evolution maps, co-occurrence networks, and strategic positioning analyses. Emergent themes were mapped to the Hornberger et al. (2023) test's 14 competencies to validate Phase 2 coverage. Phase 2 employed a cross-sectional design with Slovenian students (145 total; 40% female, 60% male; ages 7.9–21.2 years) from two secondary technical schools and one university teacher-education faculty. The validated 31-item AI literacy test (Cronbach's α=0.76) covered competencies grounded in Long & Magerko (2020) and aligned with AI4K12 big ideas—robust coverage of BI2 (representation/reasoning) and BI3 (learning), substantial BI5 (societal impact), minimal BI4 (interaction), and no BI1 (perception/robotics). Administration was proctored, device-agnostic (smartphones/tablets), single-session (35–45 min). Analysis used SPSS: 2×2 ANOVA (group×sex on total score), MANCOVA (group effect on 14 competencies, sex covariate), and MANCOVA (group effect on 4 critical-AI competencies). Effect sizes reported as partial η².
Key Findings: Bibliometric (RQ1): The field exhibited explosive growth (125% annually), moving from foundational AI concepts (2017–2022) to LLM/ChatGPT-centric integration (2023–2025). Four thematic communities emerged: (1) Core AI/ethics hub linking LLMs, pre-service teachers, assessment, responsible AI; (2) Pedagogical applications across K-12, teacher education, assessment; (3) Medical/healthcare education niche; (4) Technology acceptance/attitudes. By 2025, AI/critical-AI literacy stabilized as basic (foundational) themes; ethics remained motor/central; chatbots transitioned from motor to basic; methodologies matured (systematic reviews, SEM); K-12 and medical domains sustained high centrality. Digital literacy receded; anxiety/bias/integrity gained prominence. Instrument mapping confirmed strong BI2–BI3 and BI5 coverage, minimal BI4, absent BI1—explaining certain null findings.
Empirical (RQ2–RQ4): Teacher-education students scored modestly higher overall (M=12.40 vs. 11.97, p=0.02, η²=0.054), with significant sex effect favoring males (p=0.01, η²=0.074) and no interaction (p=0.61). Across 14 competencies (RQ3), multivariate group effect was non-significant (Wilks' λ=0.86, p=0.15), but univariate tests showed teacher-education advantages in Understanding Intelligence (p=0.002, η²=0.07, medium effect) and Programmability (p=0.045, η²=0.03, small effect). Males outperformed females on Interdisciplinarity (p=0.001, η²=0.07), Data Literacy (p=0.011, η²=0.05), and Ethics (p=0.018, η²=0.04). For critical-AI literacy (RQ4: strengths/weaknesses, human role, ethics, interdisciplinarity), no group difference emerged (Wilks' λ=0.96, p=0.314), though sex effects persisted (Interdisciplinarity p=0.001, Ethics p=0.018). Benchmarked internationally, Slovenian teacher-education students aligned with UK (IRT ≈−0.015), below German (0.38), above USA (−0.24).
Implications: Findings converge with regional evidence (Licardo et al., 2025) showing 80%+ of both cohorts lacked organized AI training, relying on informal channels—explaining competency equivalence and modest group advantages. Critical-AI parity suggests ethics/societal themes diffuse widely yet superficially without structured curricula. Recommendations: (1) embed performance-based modules (dataset audits, bias testing, prompt engineering, governance cases) in both sectors aligned to four pillars—Foundational Knowledge, Critical Appraisal, Participatory Design, Pedagogical Integration; (2) deepen teacher-education programmability into human-centered design studios; strengthen secondary students' conceptual models of intelligence and end-to-end data/ML practices; (3) close gender gaps via equity strategies (values affirmation, structured pair roles, utility-value framing, interdisciplinary transfer scaffolds); (4) develop policy frameworks spanning institutional (pillar-aligned course maps, teacher PD on LLM pedagogy) to national levels (AI literacy standards sequencing K-12→teacher education, governance/ethics mandates, equity monitoring).
Limitations: Single-site convenience sample (n=145) from Slovenia limits generalizability; uneven sex distribution constrains inference; cross-sectional design precludes causality; self-report plus BI1/BI4 coverage gaps may understate robotics/interaction competencies; modest R² (13.1% variance) indicates unmeasured factors.
Future Directions: Broaden sampling; incorporate additional databases (Scopus); extend measurement to BI1/BI4; employ mixed-methods (self-efficacy/regulation surveys, focus groups); conduct longitudinal/intervention studies testing how targeted curricula shift competencies and close gaps.
Title and Authors: "Toward a Coherent AI Literacy Pathway in Technology Education: Bibliometric Synthesis and Cross-Sectional Assessment" by Denis Rupnik (corresponding author) and Stanislav Avsec, Department for Physics and Technology, Faculty of Education, University of Ljubljana, Slovenia.
Published: November 1, 2025, in Education Sciences, 15(11), 1455. https://doi.org/10.3390/educsci15111455